This capstone project explores the application of agentic Al to improve supplier segmentation at a global healthcare company managing a complex global supplier network. The current segm_entation process relies on manual, binary yes/no assessments across 13 questions, producing outcomes that are often subjective, inconsistent across category managers, and difficult to trace back to supporting evidence. To address these challenges, we designed and developed a proof-of-concept Supplier Management Assistant powered by a multi-agent Al system. The solution integrates retrieval-augmented generation (RAG) with a two-tier data strategy, combining question-mapped primary sources with broader contextual knowledge, to automatically retrieve, synthesize, and present structured evidence for each segmentation decision. A human-in-the-loop design ensures that category managers retain final decision authority while benefiting from Al-generated recommendations grounded in existing enterprise data platforms. Expected outcomes include a reduction in segmentation time, improved decision consistency and traceability, and scalable coverage across the full supplier base. Working with the stakeholders, the project also delivers a suggested phased roadmap for scaling the solution from proof-of-concept through pilot to enterprisewide adoption, positioning the Sponsor company to transition from manual, resource-constrained supplier governance toward an intelligence-driven, Al-augmented supplier management capability.
